An Alternative Algorithm for Continuum Regression

• 발행기관: 한국자료분석학회
• 저자명: Jong-Duk Kim
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.13 No.1, 81~90쪽, 전체 10쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 2011.02.28

영문 초록

Continuum regression (CR) proposed by Stone and Brooks (1990) can generate a range of models embracing ordinary least squares regression (OLS), partial least squares regression (PLS) and principal component regression (PCR). By using a continuum coefficient, it was shown that OLS and PCR occupy the opposite ends of a continuous spectrum, with PLS lying in between somewhere. The continuum coefficient allows a smooth transition among these three regression methods. Although this provided an important idea in regression methodologies, it has mainly been discussed as theoretical perspectives and its use in practice is not prevalent. This paper provides an alternative CR algorithm in a canonical form that has good numerical properties and illustrates how to find an optimal continuum regression model in real world data.

목차

1. Introduction

2. Continuum Regression Method

3. Alternative Algorithm for Continuum Regression

4. Numerical Example

5. Conclusions

References

키워드

continuum regression continuum coefficient partial least squares regression principal component regression cross validation spectroscopic data